Predicting childhood anaemia in Ghana with explainable machine learning: A national survey analysis

Machine learning models achieved moderate predictive performance for childhood anaemia in Ghana. Gradient boosting provided the strongest discrimination, while support vector machine and logistic regression offered higher sensitivity for screening. However, these sensitivities imply that approximately 28-30% of anaemic children may be missed, which should be considered when applying these models in public health screening. Identified determinants support targeted, malaria-integrated nutrition and maternal-child interventions in high-risk groups.